Offline Time-Independent Multi-Agent Path Planning

Authors: Keisuke Okumura, François Bonnet, Yasumasa Tamura, Xavier Défago

IJCAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We present algorithms to solve OTIMAPP and demonstrate the utility of OTIMAPP via robotic applications... Section 6 shows that either PP or DBS can compute large OTIMAPP instances to some extent. Furthermore, we show that solutions keep robots moves efficient in an adverse environment for timing assumptions compared to existing approaches with runtime supports... Moreover, we demonstrate that solutions are executable with physical robots... This section empirically demonstrates that OTIMAPP solutions are computable to some extent (Sec. 6.1) and they are useful in adverse environments about timings (Sec. 6.2) through the simulation experiments. We also present OTIMAPP execution with robots (Sec. 6.3).
Researcher Affiliation Academia Keisuke Okmura , Franc ois Bonnet , Yasumasa Tamura and Xavier D efago Tokyo Institute of Technology {okumura.k, bonnet.f, tamura.y, defago.x}@coord.c.titech.ac.jp
Pseudocode Yes Algorithm 1 PP: Prioritized Planning and Algorithm 2 DBS: Deadlock-based Search.
Open Source Code Yes The appendix, code, and movie are available on https://kei18.github.io/otimapp.
Open Datasets No The paper describes generating instances on 'four-connected undirected grids picked up from [Stern et al., 2019]' and 'random graphs'. It states 'All instances were generated by setting a start si and a goal gi randomly'. However, it does not provide concrete access (URL, DOI, or specific citation to a dataset repository) for these generated instances or the random graphs used in their experiments, nor does it refer to a universally recognized public dataset by name that guarantees access.
Dataset Splits No The paper mentions running experiments on '25 identical instances' and '10 instances that OTIMAPP solutions were found by PP+' with '50 trials while changing the random seed'. However, it does not specify explicit training, validation, or test dataset splits (e.g., percentages, sample counts, or predefined splits) for the data used in the experiments.
Hardware Specification Yes The simulator was coded in C++ and the experiments were run on a desktop PC with Intel Core i9 2.8 GHz CPU and 64 GB RAM.
Software Dependencies No The paper mentions 'The simulator was coded in C++' but does not provide specific version numbers for the C++ compiler, any libraries, or other software dependencies used.
Experiment Setup No The paper describes general experimental settings such as 'timeout of 5 min' and how instances were generated (random start/goal pairs), and mentions the use of specific heuristics for the DBS solver ('descending order of the number of deadlocks with two agents'). However, it does not provide specific hyperparameter values or detailed system-level training settings like learning rates, batch sizes, or optimizer configurations that would typically be found in an 'experimental setup' section.